Learning-by-doing AI Safety Research workshop — LessWrong